Sentence matching and well-formedness

Understanding how linguistic competence controls linguistic performance is the primary aim of experimental psycholinguistic research. To attain this goal, we must interrelate three different kinds of theories: theories of the information-base (i.e., the way in which linguistic knowledge is represented), theories of the structure and operation of the language-processor itself, and theories of how the operating characteristics of the language-processor are revealed in observable behavior. Given the degrees of freedom available in each of these theories, it is difficult to find observable operating characteristics of the language processing system that have clear implications for an overall theory. To make progress, we need to discover direct correspondences between fairly general properties of the information-base and observable behavior, so that the implications of this correspondence are invariant over a reasonably wide range of assumptions in each of the problem areas. One promising correspondence in the area of syntax recognition is the accidental discovery that there is a sentence processing task that is totally insensitive to the ungrammaticality produced by violations of constraints on movement, as in example (l), but is nevertheless highly sensitive to other types of ungrammaticality, as in (2) (Freedman & Forster, 1985). This task is the sentence matching task, in which subjects have to decide as rapidly as possible whether one word string is the same as a previously presented string.